Allocation of resources for a plurality of hosts

Abstract

It is presented a method for enabling allocation of resources for a plurality of hosts. The method is performed by a server ( 1 ) and comprises identifying (S 100 ) a service running on one or more of the plurality of hosts, determining (S 140 ) a stretch factor for a recurring load pattern for the service running on the one or more of the plurality of hosts, and storing (S 150 ) the identified service together with the determined stretch factor. It is also presented a server, a computer program and a computer program product.

The invention claimed is: 
     
       1. A method for enabling allocation of resources for a plurality of hosts, the method being performed by a server and comprising:
 identifying a service running on one or more of the plurality of hosts; 
 determining a stretch factor for a recurring load pattern for the service running on the one or more of the plurality of hosts; 
 storing an identifier identifying the identified service together with the determined stretch factor, wherein the stretch factor associates a first load pattern having a first time series of a first length to a second load pattern having a second time series of a second length different from the first length; 
 using the stretch factor to predict a load pattern for the service; 
 observing a load pattern; and 
 determining whether the service caused the observed load pattern, wherein the determining comprises comparing the predicted load pattern with the observed load pattern. 
 
     
     
       2. The method of  claim 1 , further comprising:
 extracting host load data directly from the plurality of hosts, the host load data being related to the identified service. 
 
     
     
       3. The method of  claim 2 , further comprising:
 determining a load pattern for the identified service from the extracted host load data; and 
 comparing the determined load pattern with similar load patterns to define a recurring load pattern, wherein 
 the comparing comprises calculating dominant frequencies of load patterns, and 
 the dominant frequencies are calculated by a Fast Fourier Transform (FFT). 
 
     
     
       4. The method of  claim 1 , wherein the first time series and the second time series have the same key characteristics. 
     
     
       5. The method of  claim 1 , wherein the stretch factor is determined by Dynamic Time Warping (DTW). 
     
     
       6. The method of  claim 1 , wherein the stretch factor is determined by both comparing the identified load pattern with similar load patterns and comparing similar load patterns with the identified load pattern. 
     
     
       7. The method of  claim 1 , comprising:
 training a classifier with the identified service normalized with the determined stretch factor, and/or 
 training a classifier with the identified service and with the determined stretch factor. 
 
     
     
       8. The method of  claim 7 , comprising:
 predicting a load pattern for the service, utilizing the trained classifier. 
 
     
     
       9. The method of  claim 1 , wherein the plurality of hosts are a plurality of virtual hosts, and the allocation of resources is an allocation of virtual resources. 
     
     
       10. A server for enabling allocation of resources for a plurality of hosts, the server comprising:
 a processor; and 
 a computer program product storing instructions that, when executed by the processor, causes the server to: 
 identify a service running on one or more of the plurality of hosts; 
 determine a stretch factor for a recurring load pattern for the service running on the one or more of the plurality of hosts; 
 store an identifier identifying the identified service together with the determined stretch factor, wherein the stretch factor associates a first load pattern having a first time series of a first length to a second load pattern having a second time series of a second length different from the first length; 
 use the stretch factor to predict a load pattern for the service; 
 observe a load pattern; and 
 determine whether the service caused the observed load pattern by comparing the predicted load pattern with the observed load pattern. 
 
     
     
       11. The server of  claim 10 , further caused to:
 extract host load data directly from the plurality of hosts, the host load data being related to the identified service. 
 
     
     
       12. The server of  claim 11 , further caused to:
 determine a load pattern for the identified service from the extracted host load data; and 
 compare the determined load pattern with similar load patterns to define a recurring load pattern. 
 
     
     
       13. The server of  claim 12 , wherein the compare comprises calculate dominant frequencies of load patterns and the dominant frequencies are calculated by a Fast Fourier Transform (FFT). 
     
     
       14. The server of  claim 10 , wherein the first time series and the second time series have the same key characteristics. 
     
     
       15. The server of  claim 10 , wherein the stretch factor is determined by Dynamic Time Warping (DTW). 
     
     
       16. The server of  claim 10 , wherein the stretch factor is determined by both comparing the identified load pattern with similar load patterns and comparing similar load patterns with the identified load pattern. 
     
     
       17. The server of  claim 10 , further caused to:
 train a classifier with the identified service and with the determined stretch factor, and/or 
 train a classifier with the identified service normalized with the determined stretch factor. 
 
     
     
       18. The server of  claim 17 , further caused to:
 predict a load pattern for the service, utilizing the trained classifier. 
 
     
     
       19. The server of  claim 11 , wherein the plurality of hosts are a plurality of physical hosts, and the allocation of resources is an allocation of physical resources, and the host load data is physical host load data. 
     
     
       20. A computer program product comprising a computer program for enabling allocation of resources for a plurality of hosts, the computer program comprising computer program code which, when run on a server, causes the server to:
 identify a service running on one or more of the plurality of hosts; 
 determine a stretch factor for a recurring load pattern for the service running on the one or more of the plurality of hosts; 
 store an identifier identifying the identified service together with the determined stretch factor, wherein the stretch factor associates a first load pattern having a first time series of a first length to a second load pattern having a second time series of a second length different from the first length; 
 use the stretch factor to predict a load pattern for the service; 
 observe a load pattern; and 
 determine whether the service caused the observed load pattern by comparing the predicted load pattern with the observed load pattern. 
 
     
     
       21. The method of  claim 1 , further comprising:
 using the stretch factor to predict a load pattern for the service; 
 observing a load pattern; 
 determining whether the service caused the observed load pattern, wherein determining whether the service caused the observed load pattern comprises determining whether the predicted load pattern matches the observed load pattern.